Skip to content
Branch: master
Go to file
Code

Latest commit

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
src
 
 
 
 
 
 
 
 
 
 

README.md

BourGAN: Generative Networks with Metric Embeddings

Abstract:

This paper addresses the mode collapse for generative adversarial networks (GANs). We view modes as a geometric structure of data distribution in a metric space. Under this geometric lens, we embed subsamples of the dataset from an arbitrary metric space into the l2 space, while preserving their pairwise distance distribution. Not only does this metric embedding determine the dimensionality of the latent space automatically, it also enables us to construct a mixture of Gaussians to draw latent space random vectors. We use the Gaussian mixture model in tandem with a simple augmentation of the objective function to train GANs. Every major step of our method is supported by theoretical analysis, and our experiments on real and synthetic data confirm that the generator is able to produce samples spreading over most of the modes while avoiding unwanted samples, outperforming several recent GAN variants on a number of metrics and offering new features.

Publication

Chang Xiao, Peilin Zhong, and Changxi Zheng. "BourGAN: Generative Networks with Metric Embeddings." Neural Information processing systems (NIPS), 2018. (Spotlight Presentation) [PDF]

Citation

@article{xiao2018bourgan,
  title={BourGAN: Generative Networks with Metric Embeddings},
  author={Xiao, Chang and Zhong, Peilin and Zheng, Changxi},
  journal={Advances in Neural Information Processing Systems},
  year={2018}
}

Requirements

pytorch 0.4.1

numpy 1.14.3

scipy 1.1.0

matplotlib 2.2.3

The repo is still under development and will update frequently.

About

BourGAN: Generative Networks with Metric Embeddings

Resources

Releases

No releases published
You can’t perform that action at this time.